Gli effetti a breve termine del particolato atmosferico sulla salute. Un esempio di utilizzo della data Warehouse denali nell'epidemiologia ambientale

Background and aims – Short-term exposure to particulate matter (PM) has been associated with cardiorespiratory and cerebrovascular diseases. So far, most of the pertaining epidemiological studies have considered densely populated urban areas, but PM might have a different effect outside of them; furthermore, they have usually investigated outcomes such as mortality and hospitalizations and rarely drug prescriptions, that could be surrogates for adverse events that are undetectable through deaths or hospitalizations. Aim of this project was to analyze the short-term relationship between exposure to PM10 and cardiorespiratory or cerebrovascular adverse events, detected through hospitalizations and drug prescriptions, in the whole Lombardy, a highly polluted region of Northern Italy that includes areas with different geographic and meteoclimatic conditions. Materials and methods – Daily mean PM10 concentrations were provided by the environmental protection agency of Lombardy: we requested data from fixed monitoring stations for the period 2000-2008, and concentrations estimated through a Chemical Transport Model (CTM) for the period 2007-2008. Health data were retrieved from the datawarehouse DENALI, that gathers and links health administrative databases of the Lombardy Healthcare System. We extracted hospitalizations for respiratory, cardiovascular and cerebrovascular causes and drug prescriptions of cardiorespiratory treatments occurred in the population of Lombardy between 2000 and 2008. The region was split into 7 areas, which were homogeneous for pollutants emissions and meteoclimatic conditions. In each of these areas we investigated the relationship between PM and each outcome of interest using different exposure assessment and statistical methods: area-specific and municipality-specific PM10 concentrations were estimated using data from fixed monitoring stations or from the CTM; and we applied both time-series and case-crossover designs. In modeling the relationship between exposure and hospitalization we accounted for traditional confounders; while we developed a new method to account for anomalies in the time-series of the frequency of drug prescription. We evaluated the effect modification due to season, distinguishing between warm and cold season; sex; age, categorized into 2 or 3 classes; cardiorespiratory treatments before the hospitalizations, identifying subject who received a prescription during the 30/60 days before the admission; and patient complexity, evaluated through Charlson Comorbidity Index categorized into 3 classes. Area-specific estimates were finally combined through a fixed effect metanalysis. Results – Mean PM10 levels differed by area and concentration estimated through CTM data were lower than those estimated using data from monitoring stations. The study included globally 1430081 hospital admissions and 54675467 drug prescriptions. As far as the analyses of hospitalizations is concerned, the effects of PM estimated through different approaches usually differed and area-specific results were heterogeneous. For example, the metanalytic estimates of the percent variation of the risk of respiratory hospitalization relative to an increment of 10 µg/m3 in PM10 concentration distributed lag 0-1 varied from -0.37 [Confidence Interval (CI) 95%: -1.17; 0.43] to 0.34 (CI 95%: 0.15; 0.53) according to the approach. Nevertheless, in the analyses of both cardiovascular and respiratory admissions, a significant effect modification due to season emerged, regardless for the implemented method: the estimated percent variation of the risk of event was higher during the warm season. No clear pattern of effect modification appeared for the other examined variables. The probability of prescription of diuretics, betablockers, calcium channel blockers, glucocorticoids and adrenergics increased significantly with PM10 concentration. Estimated percent variations in the risk of prescriptions obtained using different models were diverse and they were usually higher in models considering exposure assessed through CTM. For respiratory drugs, we observed an effect modification due to season that was similar to that previously estimated in the analysis of hospitalizations; for some of the examined outcomes there was also evidence of an effect modification due to sex or age. Discussion – This study explored the relationship between exposure to PM10 and cardiorespiratory and cerebrovascular health in Lombardy. It highlighted the fundamental influence of the exposure assessment method on the effect estimates; furthermore it substantially confirmed the hypothesis of drug prescriptions being suitable surrogates to detect moderate adverse events. Our study has some limitations: there was possible exposure and outcome misclassification and confounding by indication, the least one applying only to the analysis of the effect modification due to treatment. However, thanks to DENALI, it was possible to investigate a rarely considered outcome, namely drug prescriptions; to involve a large unselected population and to evaluate rural and mountain areas as well as urban and suburban ones.